Artificial Intelligence Large Language Models Improve Patient Comprehension of Radiologist Magnetic Resonance Imaging Reports

Abstract

PURPOSE: To assess whether an artificial intelligence (AI) translation of a magnetic resonance imaging (MRI) report improved patient understanding of the information presented in the radiology report and to evaluate patient preferences for AI translations over traditional radiology reports. METHODS: Patients presenting to an orthopaedic surgery clinic were prospectively enrolled and randomized into 2 groups. The standard MRI group received a traditional MRI report on a multiligament knee injury written by a radiologist, whereas the AI group received an AI-translated version of the same report, generated using ChatGPT version 4. All patients completed a standardized quiz to assess comprehension of their respective reports. After the quiz, participants were provided with both reports and asked to rate their preferences between the two. Demographic information including age, sex, race, education level, area deprivation index, and orthopaedic history was collected from all patients. RESULTS: A total of 64 patients (32 in each group) with an average age of 51.9 ± 15.5 years were enrolled and randomized. No significant differences in demographic characteristics were identified between the 2 groups. Patients in the AI group scored 20% higher than those in the standard MRI group on the comprehension quiz (74.7% vs 54.7%, P < .001). Overall, 87.5% of patients preferred the AI translation whereas 4.7% preferred the standard version. Patients rated the AI translation as significantly clearer than the standard version (4.5 of 5 vs 2.2 of 5, P < .001), although they had less trust in the AI translation compared with the standard report (4.1 of 5 vs 4.5 of 5, P = .003). A higher education level was predictive of comprehension. CONCLUSIONS: AI-translated MRI reports significantly improved patient comprehension and were preferred for their clarity, despite lower trust levels compared with standard radiology reports. CLINICAL RELEVANCE: AI-translated MRI reports have the potential to enhance patient understanding of radiologic findings in orthopaedic care. However, given the low level of trust in AI-generated content observed in this study, physician oversight remains essential to ensure accuracy and sustain patient confidence.

Ähnliche Arbeiten